DE102020104292B4 - MISSING IGNITION DETECTION DEVICE FOR COMBUSTION ENGINE, MISIGNAL DETECTION SYSTEM FOR COMBUSTION ENGINE, DATA ANALYSIS DEVICE, CONTROL FOR COMBUSTION ENGINE, METHOD FOR DETECTING FAILURE AND EMERGENCY ENGINE - Google Patents

Abstract

A misfire detection device for an internal combustion engine, the misfire detection device comprising: memory means; andprocessing circuitry, wherein the storage means stores first mapping data corresponding to the case that a warm-up process is carried out in a catalyst arranged in an exhaust passage of an internal combustion engine and second mapping data corresponding to the case that the warm-up process is not carried out, both the first and second mapping data define a mapping in which a misfire variable is output using a rotational waveform variable, the misfire variable being a variable related to a probability of occurrence of misfire, the processing circuitry is adapted to execute: a detection process that detects the rotation waveform variable based on a detection value of a sensor configured to detect a rotational behavior of a crankshaft of the internal combustion engine; a determination process based on B. determines whether the misfire is occurring based on an output of the mapping using the variable as input acquired by the acquisition process, a handling process that handles occurrence of misfire by operating predetermined hardware in the case that it is determined in the determination process that the misfire has occurred, and a selection process that selects either the first mapping data or the second mapping data used in the determination process in accordance with whether the warm-up process has been performed, with an interval between the angles at which in the internal combustion engine top compression dead centers are reached, a reaching interval is, several angle intervals that are smaller than the reaching interval, several very small angular intervals are, a rotational speed of the crankshaft at each of the several very small angular intervals is an instantaneous speed and a variable that is in Is related to instantaneous speed, is an instantaneous speed variable, the rotation waveform variable is a variable that indicates a difference between multiple values of an instantaneous velocity variable or that corresponds to multiple different very short angular intervals, and the mapping returns a value of the misfire variable by performing a join operation on the Value of the rotation waveform variable and a parameter learned through machine learning.

Description

BACKGROUND

Field of invention

The present disclosure relates to a misfire detection device for internal combustion engines, a misfire detection system for internal combustion engines, a data analyzer, a controller for internal combustion engines, a method of detecting misfire in the internal combustion engine, and a reception executor.

Description of related technology

The Japanese patent application JP 2002 - 4,936 A describes, for example, an apparatus for determining the presence of misfire based on a comparison between a rotational speed difference and a determination value. Two cylinders that successively reach the top compression dead center in a chronological order are referred to as adjacent cylinders. The rotational speed difference means the difference between the rotational speed of the crankshaft resulting from the combustion stroke in one of the adjacent cylinders and the rotational speed of the crankshaft resulting from the combustion stroke in the other of the adjacent cylinders.

The determination value compared with the rotation speed difference has an appropriate value that changes depending on the operating point of the internal combustion engine and the like. This increases the manufacturing steps. Therefore, the inventor has applied a mapping using a variable which uses the rotation speed difference as an input variable. The mapping outputs a value of a misfire variable, which is a variable related to a probability that the misfire has occurred, through a join operation for the input variable and a parameter learned through machine learning. However, in this case, if the same mapping is used, regardless of whether the catalyst warm-up process is performed or not, the structure of the mapping may be complicated. This will increase the computational burden in accurately calculating the value of the misfire variable.

From the US 2014/0 261 317 A1 as well as the DE 103 45 187 B4 there are known other misfire detection devices for an internal combustion engine which can detect and correct a misfire.

Proceeding from the prior art mentioned above, the object of the present invention is to further develop a technology for detecting and correcting misfires in such a way that an increase in the computing load when determining the misfire can be prevented with a simplified configuration. This object is achieved with the features of the independent claims; advantageous developments are the subject of the subclaims.

SUMMARY OF THE INVENTION

This "Summary of the Invention" is provided to introduce, in a simplified form, a selection of concepts that are further described in the "Detailed Description" below. This "Summary of the Invention" is not intended to pinpoint key or critical features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.

Several modes and their operational effects of the present disclosure are described below.

Aspect 1. One aspect of the present disclosure is a misfire detection device for an internal combustion engine. The misfire detection means includes memory means and processing circuitry. The storage device stores first mapping data that corresponds to the case that a warm-up process is carried out in a catalytic converter which is arranged in an exhaust duct of an internal combustion engine, and second mapping data that corresponds to the case that the warm-up process is not carried out wherein each of the first and second mapping data defines a mapping in which a misfire variable is output using a rotational waveform variable. The misfire variable is a variable related to a probability of occurrence of misfire. The processing circuits are configured to carry out: a detection process that detects the rotation waveform variable that is based on a detection value of a sensor that is configured to detect a rotational behavior of a crankshaft of the internal combustion engine, a determination process that is based on an output of the mapping, that uses as input the variable detected by the detection process determines whether the misfire is occurring, a handling process that handles occurrence of misfire by operating predetermined hardware in the event that it is determined in the determination process that the misfire has occurred has occurred, and a selection process that selects either the first mapping data or the second mapping data used in the determination process in accordance with whether the warm-up process has been performed. An interval between the angles at which top compression dead centers are reached in the internal combustion engine is a reaching interval. A plurality of angular intervals smaller than the reaching interval are several very small angular intervals, a rotational speed of the crankshaft at each of the several very small angular intervals is an instantaneous speed, and a variable related to the instantaneous speed is an instantaneous speed variable. The rotation curve shape variable is a variable which indicates a difference between several values of an instantaneous speed variable or which corresponds to several different very short angular intervals. The mapping returns a value of the misfire variable by performing a join operation on the value of the rotation waveform variable and a parameter learned through machine learning.

In the embodiment described above, the input to the mapping includes the rotational waveform variable in view of the fact that the rotational behavior of the crankshaft differs at different angular intervals depending on the presence of misfire. Further, in the embodiment described above, the storage means stores a mapping separately according to whether the warm-up process is performed or not, and the processing circuit changes a mapping that calculates the value of a misfire variable depending on whether the warm-up process is performed or not. Therefore, since each mapping can be a dedicated mapping according to whether the warm-up process is performed or not, each value of the misfire variable can be calculated with high accuracy while simplifying the structure of each mapping. In the embodiment described above, therefore, the value of the misfire variable can be calculated with high accuracy while reducing the computational load compared to performing such processing with a single mapping regardless of whether the warm-up process is performed or not.

Aspect 2. The misfire detection device according to aspect 1, wherein the input of the mapping defined by the first mapping data when the warm-up process is executed includes a warm-up operation amount variable that is a variable related to the warm-up process in relation to an operation amount of an operating unit of the internal combustion engine. The detection process includes a process that detects the warm-up operation amount variable when the warm-up process is performed, and the determination process includes a process that determines the presence of the misfire based on the output of the mapping that further uses the warm-up operation amount variable determined by the detection process as the Input is captured when the warm-up process is performed.

In the embodiment described above, the value of the first misfire variable reflecting the rotational behavior of the crankshaft corresponding to the warm-up operation amount can be calculated by including the warm-up operation amount variable in the input of the mapping.

Aspect 3. The misfire detection device according to aspect 2, wherein the warm-up process includes a process that retards ignition timing compared to when the warm-up process is not performed. The warm-up operation amount variable detected by the detection process includes a variable related to an amount of retardation of the ignition timing.

Since the efficiency with which the combustion energy is converted into torque changes depending on the ignition timing, the rotational behavior of the crankshaft fluctuates depending on the ignition timing. In the embodiment described above, therefore, the value of the misfire variable, which reflects the rotational behavior of the crankshaft according to the amount of retardation of the ignition timing, can be calculated by adding a variable related to the amount of retardation of the ignition timing into the Input is included in the mapping.

Aspect 4. The misfire detection device according to aspect 2 or 3, wherein the internal combustion engine includes valve characteristic variable means that allows valve characteristics of an intake valve to be changed. The warm-up process includes a process that operates the valve characteristic variable means, and the warm-up operation amount variable detected by the detection process includes a valve characteristic variable that is a variable related to the valve characteristics.

When the valve characteristic of the intake valve is changed, the amount of overlap between the valve opening period of the intake valve and the valve opening period of the exhaust valve changes. The internal EGR amount changes according to the amount of overlap. The combustion status of the air-fuel mixture in the combustion chamber changes according to the internal one EGR amount, and consequently the rotational behavior of the crankshaft changes. In the embodiment described above, therefore, the value of the misfire variable, which reflects the rotational behavior of the crankshaft according to the amount of overlap, can be calculated by including the valve characteristic variable in the input to the mapping.

Aspect 5. The misfire detection device according to any one of Aspects 2 to 4, wherein the warm-up process includes a process that changes an air-fuel ratio of an air-fuel mixture burned in a combustion chamber of the internal combustion engine according to a progress status of the warm-up process, and the warm-up operation amount variable detected by the detection process includes an air-fuel ratio variable that is an air-fuel ratio-related variable.

When the air-fuel ratio is changed, the combustion status of the air-fuel mixture in the combustion chamber changes, and hence the rotational behavior of the crankshaft changes. In the embodiment described above, therefore, the value of the misfire variable that reflects the rotational behavior of the crankshaft according to the air-fuel ratio can be calculated by including the air-fuel ratio variable in the input to the mapping.

Aspect 6. The misfire detection device according to any one of Aspects 1 to 5, wherein the input of the mapping includes an operating point variable that is a variable that defines an operating point of the internal combustion engine. The acquisition process includes a process that acquires the operating point variable. The determination process is a process that determines the presence of the misfire based on the output of the mapping that further uses the operating point variable detected by the detection process as the input. The mapping outputs a value of the misfire variable by performing a join operation on the rotation waveform variable, the operating point variable, and the parameter learned through machine learning.

The degree to which the rotational behaviors of the crankshaft differ from each other in accordance with the presence of misfire varies according to the operating point of the internal combustion engine. Therefore, if, for example, the presence of misfire is determined based on the comparison between the difference between the current speed variables corresponding to the compression top dead center of each of the two cylinders, that is, the cylinder in which the misfire has been detected and the cylinder different from the detected cylinder , and the determination value is determined, the determination value must be adjusted for each operating point. Since the mapping that outputs the value of the misfire variable by the join operation for the rotation waveform variable, the operating point variable and the parameter learned through machine learning is the learning objective, in the embodiment described above, common parameters can be made with respect to operating points that are different from each other to be learned.

Aspect 7. One aspect of the present disclosure is a misfire detection system for an internal combustion engine. The misfire detection system includes the processing circuitry and the storage device according to any one of Aspects 1 to 5. The determination process includes an output value calculation process for calculating an output value of the mapping using a variable detected by the detection process as the input. The processing circuitry includes a first execution device and a second execution device. The first execution device is at least partially mounted in the vehicle and designed to transmit the detection process, a vehicle-side transmission process, the data that are detected by the detection process to the outside of the vehicle, a vehicle-side reception process that receives a signal on a Calculation result of the output value calculation process, and execute the handling process. The second execution device is arranged outside the vehicle and is configured to receive an external reception process that receives data transmitted by the vehicle-side transmission process, the output value calculation process, the selection process and an external transmission process that receives a signal based on a calculation result of the output value calculation process, transfers to the vehicle.

In the embodiment described above, the calculation load on the in-vehicle device can be reduced by performing the output value calculation process outside the vehicle.

Aspect 8. One aspect of the present disclosure is a data analyzer comprising the second execution device and the storage device according to the aspect 7.

Aspect 9. One aspect of the present disclosure is a controller for one Internal combustion engine. The control comprises the first execution device according to aspect 7.

Aspect 10. One aspect of the present disclosure is a misfire detection method for an internal combustion engine. The detection process, the determination process and the handling process according to any one of Aspects 1 to 6 are carried out by a computer.

According to the method described above, the same effects as above in the embodiments described in FIGS. 1 to 6 are obtained.

Figure list

• 1 Fig. 13 is a view showing an embodiment of a controller and a drive system of a vehicle according to a first embodiment;
• 2 Fig. 13 is a block diagram showing part of a process performed by the controller according to the first embodiment;
• 3 Fig. 13 is a flowchart showing a procedure of a process executed by the controller according to the first embodiment;
• 4th Fig. 13 is a timing chart showing input variables of mapping according to the first embodiment;
• 5 Fig. 13 is a flowchart showing a procedure of a process executed by the controller according to the first embodiment;
• 6 Fig. 13 is a time chart showing a waveform of rotational behavior of a crankshaft according to the first embodiment;
• 7th Fig. 13 is a block diagram showing part of a process performed by a controller according to a second embodiment;
• 8th Fig. 12 is a flowchart showing a procedure of a process executed by the controller according to the second embodiment;
• 9 Fig. 13 is a view showing a configuration of a misfire detection system according to the third embodiment; and
• 10 Fig. 13 is a flowchart showing a procedure of a process executed by the misfire detection system according to the third embodiment.

DETAILED DESCRIPTION

This description provides a thorough insight into the methods, devices, and / or systems that are described. Modifications and equivalents of the methods, devices and / or systems described will be apparent to those of ordinary skill in the art. The sequences of operations are exemplary and can be changed, as would be apparent to one of ordinary skill in the art, except for operations that necessarily occur in a particular order. Descriptions of functions and facilities that are well known to those of ordinary skill in the art may be omitted.

Exemplary embodiments can take different forms and are not limited to the examples described. However, the examples described are accurate and complete, and will convey the full scope of the disclosure to those of ordinary skill in the art.

First embodiment

Now, a first embodiment relating to a misfire detection device for an internal combustion engine will be described with reference to the drawings.

In an internal combustion engine 10 that in one in the 1 If the vehicle VC shown is mounted, a throttle valve is used 14th in an intake port 12th provided. The air coming from the intake duct 12th is sucked in, flows into the combustion chambers 18th the cylinder # 1 to # 4 if an intake valve 16 is open. Fuel gets into the combustion chamber 18th of the internal combustion engine 10 by a fuel injector 20th injected. In the combustion chamber 18th the air-fuel mixture is generated by spark discharge from an ignition device 22nd used for combustion, and the energy generated by the combustion is called rotational energy of a crankshaft 24 taken. The air-fuel mixture used for combustion is discharged into an exhaust duct 28 derived when there is an exhaust valve 26th opens. The outlet duct 28 comes with a catalyst 30th provided which has an oxygen storage capacity.

The rotational power or the torque of the crankshaft 24 is via a variable valve timing device 40 to an intake-side camshaft 42 transfer. The variable valve timing device 40 changes the relative rotational phase difference between the intake side camshaft 42 and the crankshaft 24 .

With the crankshaft 24 is a crank rotor 50 coupled with several (here 34) Tooth sections 52 which is the rotation angle of the crankshaft 24 specify. The crank rotor 50 is essentially made with tooth sections 52 provided at 10 ° CA intervals, but has a missing tooth section 54 on where the interval between adjacent tooth sections 52 30 ° CA. This is to indicate the angle of rotation, which is used as a reference for the crankshaft 24 serves.

The crankshaft 24 is mechanical with a carrier C of a planetary gear mechanism 60 connected, which forms the power split mechanism. The sun gear S of the planetary gear mechanism 60 is mechanical with the rotating shaft of a first motor generator 62 connected, and the outer gear R of the planetary gear mechanism 60 is with the rotating shaft and the drive wheel 69 a second motor generator 64 connected. An AC voltage is applied to each terminal of the first motor generator 62 from an inverter 66 is applied and an AC voltage is applied to each terminal of the second motor generator 64 from an inverter 68 created.

The control 70 controls the combustion engine 10 and operates operating units of the internal combustion engine 10 , such as the throttle valve 14th , the fuel injector 20th , the ignition device 22nd who have favourited variable valve timing device 40 and the like to control the torque, the exhaust gas component ratio and the like, which are control quantities of the internal combustion engine 10 are. The controller also controls 70 the first motor generator 62 and operates the inverter 66 to control the torque and the speed of rotation, which are the control quantities. Plus the controls 70 the second motor generator 64 and operates the inverter 68 to control the torque and the speed of rotation, which are the control quantities. The 1 shows the operating signals MS1 to MS6 of the throttle valve 14th , the fuel injector 20th , the ignition device 22nd , the variable valve timing device 40 and the inverter 66 and 68 .

When controlling the control variable, the control refers 70 on the intake air amount Ga measured by the air flow meter 80 is detected, the detection value Af of the air-fuel ratio sensor 82 that is on the upstream side of the catalyst 30th is provided, and the output signal Scr of the crank angle sensor 84 and the output signal Sca of the cam angle sensor 86 that for each angular interval between the tooth sections 52 that, with the exception of the missing tooth section 54 , every 10 ° CA are provided, output pulses. Furthermore, the control relates 70 to the coolant temperature THW, which is the temperature of the cooling coolant of the internal combustion engine 10 is that by the coolant temperature sensor 88 is detected, and the accelerator depression degree (accelerator depression degree ACCP) detected by the accelerator sensor 90 is detected.

The control 70 includes a CPU 72 , a ROM 74 , a storage device 76 which is an electrically rewritable nonvolatile memory, and a peripheral circuit 77 that communicate with each other over a local area network 78 to be able to communicate. The peripheral circuit 77 includes a circuit that generates a clock signal defining an internal operation, a power supply circuit, a reset circuit, and the like.

The control 70 performs control of the control variable by the CPU 72 running a program stored in ROM 74 is stored.

The 2 shows part of the process carried out by the CPU 72 that is implemented in the ROM 74 executes the saved program.

A calculation process M10 for the required torque is a process to obtain the required torque Trqd for the internal combustion engine 10 to be calculated to a larger value when the accelerator opening degree ACCP is large than when it is small. A target charge efficiency setting process M12 is a process to set a target charge efficiency η0 * that is used to set the torque of the internal combustion engine 10 to the required torque Trqd is required. A target charge efficiency correction process M14 is a process of calculating a target charging efficiency η * by adding the charging efficiency correction amount Δη to the target charging efficiency η0 *. A throttle actuation process M16 is a process to send the operating signal MS1 to the throttle valve 14th output so that the degree of opening of the throttle valve 14th is controlled to a larger value when the target charging efficiency η * is large than when it is small.

A basic ignition timing adjustment process M18 is a process to set the basic ignition timing aig0 which is the basic value of the ignition timing based on the rotation speed NE and the charging efficiency η which is the operating point of the internal combustion engine 10 define. The rotation speed NE is determined by the CPU 72 calculated based on the output signal Scr. Furthermore, the charging efficiency η is made by the CPU 72 is calculated based on the rotation speed NE and the intake air amount Ga. An ignition timing correction process M20 is a process to calculate the ignition timing aig by adding the ignition timing correction amount Δaig to the basic ignition timing aig0. An ignition process M22 is a process to send an operation signal MS3 to the igniter 22nd output so that the timing of the spark discharge through the igniter 22nd to the ignition timing becomes aig.

A warm-up correction process M24 comprises a process of calculating the ignition timing correction amount Δaig to a value for retarding the ignition timing aig and putting it in the ignition timing correction process M20 to be entered when an execution request for the warm-up process of the catalytic converter 30th he follows. The warm-up correction process M24 includes a process to calculate the charge efficiency correction amount Δη to a value greater than zero and put it in the target charge efficiency correction process M14 to be entered when there is a request to execute the warm-up process. When the warm-up correction process is not executed, the ignition timing correction amount Δaig and the charging efficiency correction amount Δη are set to zero. In particular, the warm-up correction process represents M24 an efficiency reduction amount vef, so that the efficiency with which the combustion energy of the air-fuel mixture in the combustion chamber is reduced 18th of the internal combustion engine 10 is converted into torque by the warm-up process, and based on this, sets the ignition timing correction amount Δaig to an amount on the retard side. The charge efficiency correction amount Δη is to increase the amount of air to achieve the required torque Trqd when the efficiency reduction amount vef is not zero. In the present embodiment, the execution request for the warm-up process is made when a logical product of the fact that the coolant temperature THW is less than or equal to a predetermined temperature and that the integrated value from the start of the intake air amount Ga is less than or equal to a predetermined value is true is.

An inlet phase difference calculation process M30 is a process to find an intake phase difference DIN that is a phase difference of a rotation angle of the intake side camshaft 42 in relation to the angle of rotation of the crankshaft 24 based on the output signal Scr from the crank angle sensor 84 and the output signal Sca of the cam angle sensor 86 to calculate. A target inlet phase difference calculation process M32 is essentially a process to determine an intake phase difference DIN * based on an operating point of the internal combustion engine 10 adjustable. In the present embodiment, the operating point is defined by the rotation speed NE and the charging efficiency η. In addition, it includes a target intake phase difference calculation process M32 a process that calculates the actual target intake phase difference DIN * with respect to the target intake phase difference DIN * according to the operating point based on the correction instruction from the warm-up correction process M24 changes when the warm-up process is in progress. In particular, the target intake phase difference calculation process includes M32 a process that changes the internal EGR amount by having an overlap period (overlap amount RO) between the valve opening periods of the intake valve 16 and the valve opening period of the exhaust valve 26th based on the correction instruction from the warm-up correction process M24 changes when the warm-up process is in progress.

An intake phase difference control process M34 is a process to send an operation signal MS4 to the variable valve timing device 40 output to the variable valve timing device 40 to operate in such a way that the inlet phase difference DIN is controlled to the target inlet phase difference DIN *.

A basic injection quantity calculation process M36 is a process of calculating a basic injection amount Qb, which is a basic value of the fuel amount, based on a charge efficiency η to be the air-fuel ratio of the air-fuel mixture in the combustion chamber 18th to bring on the target air-fuel ratio. In particular, when the charging efficiency η is expressed as a percentage, for example, the basic injection amount calculation process can be used M36 be a process of calculating the basic injection amount Qb by multiplying the charge efficiency η by the fuel amount QTH for every 1% of the charge efficiency η to bring the air-fuel ratio to the target air-fuel ratio. The basic injection amount Qb is an amount of fuel that is calculated to make the air-fuel ratio based on the amount of air entering the combustion chamber 18th is loaded to control the target air-fuel ratio. In the present embodiment, a theoretical air-fuel ratio is illustrated as a target air-fuel ratio.

A feedback process M40 is a process to calculate a feedback correction coefficient KAF obtained by adding "1" to the correction ratio δ of the basic injection amount Qb, which serves as a feedback operation amount, which is an operation amount for feedback controlling the detection value Af to the target value Af * is, and output this. In particular, the feedback process represents M40 a sum of each output value of a proportional element and a differentiation element having as an input the difference between the detection value Af and the target value Af *, and an output value of an integration element holding and outputting an integrated value of a value corresponding to the difference as a Correction ratio δ a.

A calculation process M42 for the required injection amount is a process of calculating a required injection amount Qd by multiplying the basic injection amount Qb by the feedback coefficient KAF.

An injector actuation process M44 is a process to send an operating signal MS2 to the fuel injector 20th output, so that fuel corresponding to the required injection amount Qd within a combustion stroke from the fuel injection valve 20th is injected.

A goal setting process M46 is a process to set the target value Af *. The target setting process M46 includes a process that, in response to a command from the warm-up correction process M24 , during the execution of the warm-up process, the target value Af * in the first half of the warm-up process is leaner than a stoichiometric point Afs corresponding to the theoretical air-fuel ratio, and that the target value Af * in the second half of the warm-up process is richer than the stoichiometric point Afs . This is a setting that takes into account that it is difficult to purge the unburned fuel before the catalytic converter 30th is heated.

The control 70 performs a process that detects the presence of misfires during the period of operation of the internal combustion engine 10 certainly. In this time, in view of the fact that the control changes largely depending on whether the warm-up process is being carried out or not, the presence of misfires is determined by different processes.

The 3 Fig. 10 shows a procedure of a process related to misfire detection. The Indian 3 The process shown is done by the CPU 72 that implemented the misfire program 74a that is in the ROM 74 is stored, executes repeatedly, for example, at a predetermined rate. Below, the step number of each process is represented by the number with an "S" in front of it.

In the sequence of processes included in the 3 is determined by the CPU 72 first whether the warm-up process is running ( S8 ). When it is determined that the warm-up process is not running ( S8 : YES), the CPU detects 72 the short rotation time T30 ( S10 ). The short rotation time T30 is set by the CPU 72 based on the output signal Scr of the crank angle sensor 84 determined by measuring the amount of time it takes for the crankshaft to run 24 rotated by 30 ° CA. Next up is the CPU 72 the newest short rotation time T30, which is in the process of S10 has been detected as the short rotation time T30 (0), and the variable "m" of the short rotation time T30 (m) is set to a larger value as the earlier value increases ( S12 ). That is, since “m = 1, 2, 3, ...”, the short rotation time becomes T30 (m-1) immediately before performing the process of S12 defined as the short rotation time T30 (m). Thus, for example, the short rotation time T30 obtained by the process of S10 has been captured as the process of 3 The last time it was executed was the short rotation time T30 (1).

Next, the CPU determines 72 whether those in the process of S10 The detected short rotation time T30 is a period of time required for the rotation of an angular interval of 30 ° CA from the compression top dead center to the compression top dead center of one of the cylinders # 1 to # 4 ( S14 ). When determining the period of time required for the rotation of the angular interval to the compression top dead center (S14: YES), the CPU calculates 72 first the value of the rotation waveform variable so that it is the input to the determination process that determines the presence of misfire to determine the presence of misfire of a cylinder that reached compression top dead center 360 ° CA earlier.

That is, the CPU 72 first calculates the difference between the 180 ° apart values of the short rotation time T30 with respect to the angular interval from 30 ° CA before the compression top dead center to the compression top dead center as the inter-cylinder variable ΔTa (S16). In particular, the CPU 72 the intermediate cylinder variable ΔTa (m-1) to "T30 (6m-6) - T30 (6m)", where "m = 1, 2, 3, ...".

The 4th illustrates the inter-cylinder variable ΔTa. In the present embodiment, compression top dead center occurs in the order of cylinder # 1, cylinder # 3, cylinder # 4, and cylinder # 2, and the combustion stroke is illustrated in that order. The 4th FIG. 10 shows an example in which the misfire presence detection target is the # 1 cylinder by setting the short rotation time T30 (0) of the angular interval from 30 ° CA before compression top dead center to compression top dead center of cylinder # 4 in the process of S10 is captured. In this case, the inter-cylinder variable ΔTa (0) is a difference between the short rotation times T30 corresponding to both the compression top dead center of cylinder # 4 and the compression top dead center of cylinder # 3, which has previously reached compression top dead center. In the 4th the inter-cylinder variable ΔTa (2) is described as the difference between the short rotation time T30 (12) corresponding to the compression top dead center of cylinder # 1 at which misfire is to be detected and the short rotation time T30 (18) corresponding to the compression top dead center of cylinder # 2.

Back to 3 : The CPU 72 calculates the inter-cylinder variable ΔTb, which is the difference between values of the inter-cylinder variables ΔTa (0), ΔTa (1), ΔTa (2), ... that are 720 ° CA apart ( S18 ). In particular, the CPU 72 the intermediate cylinder variable ΔTb (m-1) to "ΔTa (m-1) - ΔTa (m + 3)", where "m = 1, 2, 3, ...".

The 4th illustrates the inter-cylinder variable ΔTb. The 4th shows that the inter-cylinder variable ΔTb (2) is “ΔTa (2) - Ta (6)”.

Back to 3 : The CPU 72 calculates the fluctuation pattern variable FL indicating the relative relationship between the inter-cylinder variable ΔTb corresponding to the cylinder in which misfire is to be detected and the inter-cylinder variable ΔTb corresponding to the other cylinders ( S20 ). In the present embodiment, the fluctuation pattern variables FL [02], FL [12], FL [32] are calculated.

Here, the fluctuation pattern variable FL [02] is defined by “ΔTb (0) / ΔTb (2)”. That is, using the example of 4th : the fluctuation pattern variable FL [02] is a value obtained by dividing an inter-cylinder variable ΔTb (0) corresponding to the cylinder # 4 that reaches the compression top dead center after the next time by the inter-cylinder variable ΔTb (2) corresponding to the cylinder # 1 at misfire is to be detected is determined. The fluctuation pattern variable FL [12] is defined by "ΔTb (1) / ΔTb (2)". That is, using the example of 4th : the fluctuation pattern variable FL [12] is a value obtained by dividing an inter-cylinder variable ΔTb (1) corresponding to the cylinder # 3 that will reach compression top dead center by the inter-cylinder variable ΔTb (2) corresponding to the cylinder # 1 where Misfire is to be detected is determined. The fluctuation pattern variable FL [32] is defined by "ΔTb (3) / ΔTb (2)". That is, using the example of 4th : the fluctuation pattern variable FL [32] is a value detected by dividing an inter-cylinder variable ΔTb (3) corresponding to the cylinder # 2 that has reached compression top dead center by the inter-cylinder variable ΔTb (2) corresponding to the cylinder # 1 at the misfire is to be determined.

Next, the CPU captures 72 the rotation speed NE and the charge efficiency η, which are the operating point of the internal combustion engine 10 define ( S22 ).

Then the CPU resets 72 the values of the rotational waveform variables generated by the processes of S18 , S20 have been recorded and the value of the variable generated by the process of S22 has been detected, in place of the input variables x (1) to x (6) of the mapping that outputs the misfire variable PR, which is a variable related to the probability that the misfire to be detected has occurred in the cylinder ( S24 ). That is, the CPU 72 sets the intermediate cylinder variable ΔTb (2) in place of the input variable x (1), sets the fluctuation pattern variable FL [02] in the place of the input variable x (2), sets the fluctuation pattern variable FL [12] in the place of the input variable x (3) and sets the fluctuation pattern variable FL [32] in place of the input variable x (4). Furthermore, the CPU 72 the rotation speed NE in the place of the input variable x (5) and sets the charge efficiency η in the place of the input variable x (6).

Next the CPU calculates 72 the value of the misfire variable PR, which is the output value of the mapping, by entering the input variables x (1) through x (6) into the mapping obtained by the post-warm-up mapping data 76a which is defined in the in the 1 storage device shown 76 are stored ( S26 ).

In the present embodiment, this mapping is formed by a neural network that includes an intermediate layer. The neural network comprises the input-side coefficient wA (1) jk (j = 0 to n, k = 0 to 6) and the activation function h (x), which serves as an input-side non-linear mapping that maps each of the outputs of the input-side linear mapping, which is the linear mapping defined by the input-side coefficient wA (1) jk, converts nonlinearly. In the present embodiment, ReLU is shown as the activation function h (x) by way of example. Here, wA (1) j0 and the like are bias parameters, and the input variable x (0) is defined as “1”.

The neural network further includes the output-side coefficient wA (2) ij (i = 1 to 2, j = 0 to n) and the softmax function that outputs the misfire variable PR using each of the prototype variables yR (1) and yR (2) or which outputs the output-side linear mapping, which is a linear mapping defined by the output-side coefficient wA (2) ij, as inputs. Thus, in the present embodiment, the misfire variable PR is obtained by quantifying the probability that the misfire actually occurred as a continuous value within a predetermined range that is larger than “0” and smaller than “1”.

Next, the CPU determines 72 whether the value of the misfire variable PR is greater than or equal to the determination value Pth ( S28 ). When it is determined that the value is greater than or equal to that Determination value Pth is (S28: YES), the CPU increments 72 the counter CR ( S30 ). Then the CPU determines 72 whether from a point in time when the process of S28 was run for the first time, or from a point in time the process was run from S36 which will be described later has been performed, or a predetermined period of time has passed ( S32 ). Here, the predetermined period of time is longer than the period of one combustion stroke. The predetermined period of time may be ten times a combustion stroke or more.

When it is determined that the predetermined time has elapsed (S32: YES), the CPU determines 72 whether the counter CR is greater than or equal to the threshold value Cth ( S34 ). This process is a process that determines whether or not misfires have occurred with a frequency exceeding the allowable range. When it is determined that the value is smaller than the threshold value Cth (S34: NO), the CPU initializes 72 the counter CR ( S36 ). In contrast, when it is determined to be greater than or equal to the threshold value Cth (S36: YES), the CPU performs 72 a notification process that includes the information in the 1 shown warning light 100 operates to encourage the user to handle the anomaly ( S38 ).

When the processes of S36 and S38 are completed or if in the processes of S8 , S14 , S28 , S32 a negative determination is made, the CPU terminates 72 first of all the in the 3 shown sequence of processes.

The 5 Fig. 13 shows a procedure of a process related to the detection of misfire. The Indian 5 The process shown is done by the CPU 72 that implemented the misfire program 74a that is in the ROM 74 is stored, executes repeatedly, for example, at a predetermined rate. In the 5 processes that correspond to the 3 processes shown are labeled with the same step number for convenience.

In the sequence of processes included in the 5 shown, the CPU runs 72 with the process of S10 when it is determined that the warm-up process is being carried out (S8: NO), and thereafter acquires an efficiency reduction amount vef, a target value Af * and an overlap amount RO in addition to the rotation speed NE and the charging efficiency η when the process of FIG S20 connected ( S22a ).

Next the CPU resets 72 the value of the processes of S18 , S20 , S22a recorded variables in place of the input variables x (1) to x (6) ( S24a ). That is, the CPU 72 sets the same variables as the process of S24 in place of the input variables x (1) to x (6) and also sets the efficiency reduction scope vef in place of the input variable x (7), sets the target value Af * in place of the input variable x (8) and sets the extent of overlap RO the position of the input variable x (9).

Next the CPU calculates 72 the value of the misfire variable PR, which is the output value of the mapping by inputting the input variables x (1) through x (9) into the mapping determined by the warm-up mapping data 76b that are in the in the 1 storage device shown 76 are stored, is defined ( S26a ).

In the present embodiment, this mapping is formed by a neural network that includes an intermediate layer. The neural network comprises the input-side coefficient wB (1) jk (j = 0 to n, k = 0 to 9) and the activation function h (x), which serves as an input-side non-linear mapping that maps each of the outputs of the input-side linear mapping, which is the linear mapping defined by the input-side coefficient wB (1) jk, converts nonlinearly. In the present embodiment, ReLU is shown as the activation function h (x) by way of example. Here, wB (1) j0 and the like are bias parameters, and the input variable x (0) is defined as “1”.

Furthermore, the neural network includes the output-side coefficient wB (2) ij (i = 1 to 2, j = 0 to n) and the softmax function which outputs the misfire variable PR using each of the prototype variables yR (1) and yR (2) or that outputs the output-side linear mapping, which is a linear mapping defined by the output-side coefficient wB (2) ij, as inputs.

When the process of S26a is complete, the CPU performs 72 the processes after S28 out.

The post-warm-up mapping data 76a are generated in the following manner, for example. That is, the internal combustion engine 10 is put to the test with the one with the crankshaft 24 connected dynamometer is operated, and the fuel injection is operated at a timing randomly selected from the timings at which the fuel required in each of cylinders # 1 to # 4 should be injected after the engine has been warmed up 10 stopped. In cylinders where the fuel injection is stopped, the data with a value of the misfire variable PR as "1" was used as teacher data, and in cylinders where the fuel injection is not stopped, the data with a value of the misfire variable PR as " 0 “included in the teacher data. Then the value of the misfire variable PR becomes through the same process as the processes of S24 and S26 using the rotational waveform variable for each point in time and the value of the process of S22 calculated variables. The values of the input side coefficient wA (1) jk and the output side coefficient wA (2) ij are learned so as to reduce the difference between the thus calculated value of the misfire variable PR and the teacher data. In particular, the values of the input-side coefficient wA (1) jk and the output-side coefficient wB (1) ij can be learned in order to minimize the tolerance entropy.

In the warm-up mapping data 76b the above processes can be changed to the value of the misfire variable PR by the same processes as the processes of S24a and S26a in the warm-up period, and the values of the input-side coefficient wB (1) jk and the output-side coefficient wB (2) ij can be learned.

As described above, by using machine learning, the post-warm-up mapping data 76a and the warm-up mapping data 76b can be learned using the teacher data that is generated by the internal combustion engine 10 is operated relatively freely, while different operating points are taken. Compared to the case that the map data for each operating point is based on the detection of the behavior of the crankshaft 24 can be adjusted in the presence of misfire, the manufacturing steps can thus be reduced.

The operation and effect of the present embodiment will now be described.

The CPU 72 determines the presence of misfire by calculating the value of the misfire variable PR based on the rotation waveform variable. Here the CPU calculates 72 the value of the misfire variable PR using the post-warm-up mapping data 76a if the warm-up process is not running, and the CPU 72 calculates the value PR of the misfire variable PR using the warm-up mapping data 76b when the warm-up process is in progress. The behavior of the crankshaft differs when the catalytic converter is warmed up 24 from the case that the warm-up process is not performed, for example, the internal combustion engine 10 is operated with reducing the combustion efficiency.

In the 6 the transition of the short rotation time T30 to the normal period is shown by a broken line, the transition of the micro rotation period T30 when the warm-up process is not performed, when a misfire occurs is shown by a solid line, and the transition of the short rotation time T30 when the warm-up process executed when a misfire occurs is shown by a chain line. Like in the 5 as shown, in the case that the warm-up process is performed when a misfire has occurred, the fluctuation in the short rotation time T30 becomes smaller than when the warm-up process is not performed. Therefore, the difference between the fluctuation in the short rotation time T30 when the warm-up process is performed when the misfire has not occurred and the fluctuation in the short rotation time T30 when the warm-up process is not performed when the misfire has occurred becomes small. Thus, if the process is not changed according to the presence of the warm-up process, there is a possibility that the accuracy of identification between the fluctuation in the short rotation time T30 when the warm-up process is performed when no misfire has occurred and the fluctuation in the short rotation time T30 when the warm-up process is not performed when the misfire has occurred.

In the present embodiment, therefore, the post-warm-up mapping data 76a and the warm-up mapping data 76b set as different dates. If the same data is used regardless of whether the warm-up process is in progress or not, a requirement to increase the size of the input variable or a requirement to increase the number of intermediate layers may arise, and the mapping structure tends to be complex. In the present embodiment, the mapping can be simplified by using different mappings depending on whether the warm-up process is being carried out or not, and as a result, the calculation load can be reduced while the value of the misfire variable PR is calculated with high accuracy.

1. (1) Die Rotationsgeschwindigkeit NE und die Ladungseffizienz η, die als Betriebspunktvariablen dienen, die den Betriebspunkt des Verbrennungsmotors 10 definieren, werden als Eingaben des Mappings verwendet. Der Betriebsumfang der Betriebseinheit des Verbrennungsmotors 10, wie zum Beispiel des Kraftstoffeinspritzventils 20 oder der Zündeinrichtung 22, wird tendenziell auf Basis des Betriebspunkts des Verbrennungsmotors 10 bestimmt. Daher ist die Betriebspunktvariable eine Variable, die Informationen umfasst, die in Beziehung zum Betriebsumfang jeder Betriebseinheit stehen. Durch Verwenden der Betriebspunktvariablen als Eingabe des Mappings kann daher der Wert der Fehlzündungsvariablen PR auf Basis der Informationen in Bezug auf den Betriebsumfang jeder Betriebseinheit berechnet werden, und als ein Ergebnis kann der Wert der Fehlzündungsvariablen PR mit höherer Genauigkeit berechnet werden, indem eine Änderung im Rotationsverhalten der Kurbelwelle 24 durch den Betriebsumfang reflektiert wird.

The present embodiment described above also has the following operational effects.

1. (1) The rotation speed NE and the charge efficiency η, which serve as operating point variables that represent the operating point of the internal combustion engine 10 are used as inputs for the mapping. The scope of operation of the operating unit of the internal combustion engine 10 such as the fuel injector 20th or the ignition device 22nd , tends to be based on the operating point of the internal combustion engine 10 certainly. Therefore, the operating point variable is a variable that includes information related to the Operational scope of each operational unit. By using the operating point variable as the input of the mapping, therefore, the value of the misfire variable PR can be calculated based on the information relating to the amount of operation of each operating unit, and as a result, the value of the misfire variable PR can be calculated with higher accuracy by changing a change in the rotational behavior the crankshaft 24 is reflected by the scope of operation.

• (2) Der Effizienzreduzierungsumfang vef ist in der Eingabevariablen eingeschlossen. Somit können ausführlichere Informationen zur Wirkung auf das Rotationsverhalten der Kurbelwelle 24 als im Vergleich zu dem Fall ermittelt werden, dass die Binärvariable, die angibt, ob der Warmlaufprozess ausgeführt wird oder nicht, als eine Eingabevariable verwendet wird, wobei der Wert der Fehlzündungsvariable PR mit höherer Genauigkeit einfacher berechnet werden kann.
• (3) Der Überlappungsumfang RO ist in der Eingabevariablen eingeschlossen. Die interne AGR-Menge unterscheidet sich abhängig vom Überlappungsumfang RO, der Verbrennungsstatus des Luft-Kraftstoff-Gemischs in der Brennkammer 18 ändert sich abhängig von der internen AGR-Menge, und folglich ändert sich das Rotationsverhalten der Kurbelwelle 24. In der vorliegenden Ausführungsform kann daher der Wert der Fehlzündungsvariablen PR, der das Rotationsverhalten der Kurbelwelle 24 entsprechend dem Überlappungsumfang RO reflektiert, berechnet werden, indem der Überlappungsumfang RO in die Eingabe in das Mapping eingeschlossen wird.
• (4) Der Zielwert Af* ist in der Eingabevariablen eingeschlossen. Wenn sich das Luft-Kraftstoff-Verhältnis ändert, ändert sich der Verbrennungsstatus des Luft-Kraftstoff-Gemischs in der Brennkammer 18, und folglich ändert sich das Rotationsverhalten der Kurbelwelle 24. In der vorliegenden Ausführungsform kann daher der Wert der Fehlzündungsvariablen PR, der das Rotationsverhalten der Kurbelwelle 24 entsprechend dem Luft-Kraftstoff-Verhältnis reflektiert, berechnet werden, indem der Zielwert Af* in die Eingabe in das Mapping eingeschlossen wird.
• (5) Die Rotationskurvenformvariable, welche die Eingabevariable x werden soll, wird generiert, indem ein Wert in der Nähe des oberen Verdichtungstotpunkts in der Kurzrotationszeit T30 selektiv verwendet wird. Die Differenz, die beim Vorhandensein von Fehlzündung am häufigsten auftritt, ist der Wert in der Nähe des oberen Verdichtungstotpunkts in der Kurzrotationszeit T30. Die zum Bestimmen des Vorhandenseins von Fehlzündung nötigen Informationen können daher so weit wie möglich erfasst werden, während unterbunden wird, dass sich die Größe der Eingabevariablen x erhöht, indem der Wert in der Nähe des oberen Verdichtungstotpunkts in der Kurzrotationszeit T30 selektiv verwendet wird.
• (6) Die Zwischenzylindervariable ΔTb(2) ist in der Rotationskurvenformvariablen eingeschlossen. Die Zwischenzylindervariable ΔTb(2) wird ermittelt, indem im Voraus die Differenz zwischen der Kurzrotationszeit T30 entsprechend dem oberen Verdichtungstotpunkt zwischen dem Zylinder, bei dem Fehlzündung detektiert werden soll, und dem dazu benachbarten Zylinder eindimensional quantifiziert wird. Die zum Bestimmen des Vorhandenseins von Fehlzündung nötigen Informationen können daher mit einer Variablen mit einer geringen Anzahl an Größen effizient erfasst werden.
• (7) Die Rotationskurvenformvariable schließt nicht nur die Zwischenzylindervariable ΔTb(2), sondern auch die Schwankungsmustervariablen FL ein. Da Vibrationen von der Straßenoberfläche und dergleichen an der Kurbelwelle 24 überlagert werden, kann eine fehlerhafte Bestimmung auftreten, wenn die Rotationskurvenformvariable nur die Zwischenzylindervariable ΔTb(2) ist. In der vorliegenden Ausführungsform wird der Wert der Fehlzündungsvariablen PR unter Verwendung der Schwankungsmustervariablen FL zusätzlich zur Zwischenzylindervariable ΔTb(2) berechnet, wobei der Wert der Fehlzündungsvariablen PR auf einen Wert eingestellt werden kann, der den Grad (die Wahrscheinlichkeit) der Wahrscheinlichkeit von Fehlzündung genauer als für den Fall angibt, dass der Wert der Fehlzündungsvariablen nur anhand der Zwischenzylindervariable ΔTb(2) berechnet wird.

Further, by using the operating point variable as an input variable, the value of the misfire variable PR is calculated by the join operation of the input side coefficients wA (1) jk, wB (1) jk, which are machine learning-learned parameters of the rotational waveform variable and the operating point variable. There is thus no need to adapt the adaptation value for each operating point variable. On the other hand, when the inter-cylinder variable ΔTb and the determination value are compared, for example, there is no need to adjust the determination value for each operating point variable. This increases the manufacturing steps.

• (2) The efficiency reduction amount vef is included in the input variable. This allows more detailed information on the effect on the rotational behavior of the crankshaft 24 can be determined as compared to the case where the binary variable indicating whether or not the warm-up process is being performed is used as an input variable, and the value of the misfire variable PR can be calculated more easily with higher accuracy.
• (3) The amount of overlap RO is included in the input variable. The internal EGR amount differs depending on the amount of overlap RO, the combustion status of the air-fuel mixture in the combustion chamber 18th changes depending on the internal EGR amount, and consequently the rotational behavior of the crankshaft changes 24 . In the present embodiment, therefore, the value of the misfire variable PR, which is the rotational behavior of the crankshaft 24 according to the amount of overlap RO, can be calculated by including the amount of overlap RO in the input in the mapping.
• (4) The target value Af * is included in the input variable. When the air-fuel ratio changes, the combustion status of the air-fuel mixture in the combustion chamber changes 18th , and consequently the rotational behavior of the crankshaft changes 24 . In the present embodiment, therefore, the value of the misfire variable PR, which is the rotational behavior of the crankshaft 24 according to the air-fuel ratio, can be calculated by including the target value Af * in the input in the mapping.
• (5) The rotation waveform variable to be the input variable x is generated by selectively using a value near the compression top dead center in the short rotation time T30. The difference that occurs most frequently in the presence of misfire is the value near compression top dead center in the short rotation time T30. The information necessary to determine the presence of misfire can therefore be grasped as much as possible while preventing the size of the input variable x from increasing by selectively using the value near compression top dead center in the short rotation time T30.
• (6) The inter-cylinder variable ΔTb (2) is included in the rotational curve shape variable. The inter-cylinder variable ΔTb (2) is obtained by one-dimensionally quantifying in advance the difference between the short rotation time T30 corresponding to the compression top dead center between the cylinder at which the misfire is to be detected and the cylinder adjacent thereto. The information necessary to determine the presence of misfire can therefore be efficiently acquired with a variable having a small number of sizes.
• (7) The rotation waveform variable includes not only the inter-cylinder variable ΔTb (2) but also the fluctuation pattern variable FL. Because vibrations from the road surface and the like on the crankshaft 24 are superimposed, an erroneous determination may occur when the rotation waveform variable is only the inter-cylinder variable ΔTb (2). In the present embodiment, the value of the misfire variable PR is calculated using the fluctuation pattern variable FL in addition to the inter-cylinder variable ΔTb (2), and the value of the misfire variable PR can be set to a value more accurate than the degree (probability) of the probability of misfire indicates the case that the value of the misfire variable is calculated from only the inter-cylinder variable ΔTb (2).

In the present embodiment, the value of the misfire variable PR is additionally determined by the join operation of the inter-cylinder variable ΔTb (2) and the fluctuation pattern variable FL by the input-side coefficients wA (1) jk, wB (1) jk learned by machine learning Parameters are calculated. Compared to the case where the presence of misfire is determined based on the comparison between the inter-cylinder variable ΔTb (2) and the determination value and the comparison between the fluctuation pattern variable FL and the determination value, the presence of misfire can be determined based on a more precise relationship of the inter-cylinder variables ΔTb (2) and the fluctuation pattern variable FL and the presence of misfire can be determined.

Second embodiment

A second embodiment will be described below with reference to the drawings with a focus on the differences from the first embodiment.

The 7th shows part of a process carried out by the controller 70 is carried out according to the present embodiment. The Indian 7th The process shown is done by the CPU 72 implemented that in the ROM 74 executes the saved program. In the 7th processes that correspond to the 2 processes shown correspond to the same reference numerals for the sake of convenience.

An amplitude value variable output process M50 is a process of calculating an amplitude value variable α of dither control different from the air-fuel ratio of the air-fuel mixture burned in the cylinders and outputting it while the air-fuel ratio is set when the air-fuel mixture in each of cylinders # 1 to # 4, which is the air-fuel mixture in a period in which the crankshaft 24 rotated twice, combined into one as the target value Af *. Here, in the dither control according to the present embodiment, one of the first cylinder # 1 to the fourth cylinder # 4 is a rich combustion cylinder in which the air-fuel ratio of the air-fuel mixture is richer than the theoretical air-fuel ratio. Ratio is, and the remaining three cylinders are lean-burn cylinders in which the air-fuel ratio of the air-fuel mixture is leaner than the theoretical air-fuel ratio. Then, the injection amount in the rich cylinder is set to "1 + α" times the required injection amount Qd, and the injection amount in the lean cylinders becomes "1 - (α / 3)" times the required injection amount Qd set. Thus, if the amount of air charged into each cylinder # 1 to # 4 in one combustion stroke is the same, the following two values (A) and (B) are equal to each other.

Value (A): Sum (here "α" itself) for the number of occurrences (here once) in the combustion stroke of the cylinder for rich combustion in the period in which the crankshaft rotates twice, in relation to the increase ratio (here "α") to the required injection quantity Qd in the cylinder for rich combustion.

Value (B): Sum (here "α" itself) for the number of occurrences (here three times) in the combustion stroke of the cylinder with lean combustion in the period in which the crankshaft rotates twice, of the reduction ratio (here "α / 3") in relation to the required injection amount Qd in the lean-burn cylinder.

If the amount of air charged into each of cylinders # 1 to # 4 is the same in one combustion stroke by making the value (A) and value (B) equal to each other, the air-fuel ratio can be made the same the target value Af * can be made when the air-fuel mixture contained in each of cylinders # 1 to # 4 of the internal combustion engine 10 is burned, is combined into one.

In the period of the warm-up process, the amplitude value variable α becomes through the amplitude value variable output process M50 set to a value greater than zero. In particular, the amplitude value includes variable outputting process M50 a process to variably set the amplitude value variable α based on the rotation speed NE and the charging efficiency η. In particular, the amplitude value variable α is set by the CPU 72 is calculated with map data in a state in which the map data including the rotation speed NE and the charging efficiency η as input variables and the amplitude value variable α as an output variable are in the ROM 74 are stored in advance. The 7th shows that the amplitude value variable α is zero in a region where the rotation speed NE and the charging efficiency η are large. This is due to the fact that it is considered that in a high load area or the like, the energy flow rate of the exhaust gas entering the catalyst increases 30th flows even when dither control is not performed.

The map data is set data of a discrete value of the input variable and a value of the output variable corresponding to each value of the input variable. The map calculation can be, for example, a process that takes the value of the output variable of the corresponding map data as a Has calculation result when the value of the input variable coincides with one of the values of the input variables of the map data, and which has a value obtained by interpolating the values of a plurality of output variables included in the map data as a calculation result when the value of the Input variables does not match one of the values of the input variables of the map data.

A correction coefficient calculation process M52 is a process to calculate the required injection quantity correction coefficient Qd for the rich combustion cylinder by adding the amplitude value variable α to “1”. A dither correction process M54 is a process to calculate the target injection amount Q * of cylinder # w, which should be the rich combustion cylinder, by multiplying the required injection amount Qd by the correction coefficient “1 + α”. Here “w” is one from “1” to “4”.

A multiplication process M56 is a process to multiply the amplitude value variable α by “-1/3” and a correction coefficient calculation process M58 is a process to calculate the required injection amount correction coefficient Qd for the lean-burn cylinder by taking the output value of the multiplication process M56 is added to "1". A dither correction process M60 is a process to calculate a target injection amount Q * of the # x, # y, # z cylinders, which should be the lean-burn cylinders, by multiplying the required injection amount Qd by the correction coefficient “1- (α / 3)” becomes. Here “x”, “y”, “z” are one from “1” to “4”, and “w”, “x”, “y”, “z” are different from each other.

An injector actuation process M44 outputs an operating signal MS2 to the fuel injection valve 20th of the cylinder # w which should be the rich combustion cylinder based on the injection amount target value Q * obtained by the dither correcting process M54 is output and represents the total amount of fuel coming out of the fuel injector 20th is to be injected as a quantity corresponding to the injection quantity target value Q *. There is also the fuel injector actuation process M44 an operating signal MS2 to the fuel injection valve 20th of cylinders # x, # y, # z, which should be the lean-burn cylinders, based on the injection amount target value Q * obtained by the dither correction process M60 is output and represents the total amount of fuel coming out of the fuel injector 20th is to be injected as a quantity corresponding to the injection quantity target value Q *.

The 8th Fig. 13 shows a procedure of a process related to the detection of misfire. The Indian 8th The process shown is done by the CPU 72 that implemented the misfire program 74a that is in the ROM 74 is stored, executes repeatedly, for example, at a predetermined rate. In the 8th processes that correspond to the 5 processes shown are labeled with the same step number for convenience.

In the in the 8th The sequence of processes shown is recorded by the CPU 72 when the process of S20 is completed, the amplitude value variable α in addition to the rotational speed NE and the charge efficiency η ( S22b ).

Next the CPU resets 72 the value of the processes of S18 , S20 , S22b recorded variable in place of the input variable x ( S24b ). That is, the CPU 72 sets the value of the same variable as the process of S24 takes the place of the input variables x (1) to x (6) and sets the amplitude value variable α in place of the input variable x (7).

Next up is the CPU 72 Enter the input variables x (1) to x (7) into the mapping that is generated by the warm-up mapping data 76b is defined to calculate the value of the misfire variable PR, which is the output value of the mapping ( S26b ).

In the present embodiment, this mapping is formed by a neural network that includes an intermediate layer. The above-mentioned neural network comprises the input-side coefficient wB (1) jk (j = 0 to n, k = 0 to 7) and the activation function h (x) serving as the input-side non-linear mapping that each of the outputs of the input-side linear Mappings, which is the linear mapping defined by the input-side coefficient wB (1) jk, converts non-linearly. In the present embodiment, ReLU is shown as the activation function h (x) by way of example. Here, wB (1) j0 and the like are bias parameters, and the input variable x (0) is defined as “1”.

Furthermore, the neural network includes the output-side coefficient wB (2) ij (i = 1 to 2, j = 0 to n) and the softmax function which outputs the misfire variable PR using each of the prototype variables yR (1) and yR (2) or that outputs the output-side linear mapping, which is a linear mapping defined by the output-side coefficient wB (2) ij, as inputs.

When the process of S26b is completed, the CPU shuts down 72 with the processes S28 away.

Third embodiment

A third embodiment will be described below with reference to the drawings with a focus on the differences from the first embodiment.

In the present embodiment, the calculation process for the misfire variable PR is performed outside the vehicle.

The 9 Fig. 10 shows a misfire detection system according to the present embodiment. In the 9 elements that match the ones in the 1 The elements shown correspond to the same reference numerals for the sake of convenience.

The ones in the 9 shown control 70 in the vehicle VC comprises a communication device 79 . The communication facility 79 is a device for communicating with a control center 120 over the network 110 outside the vehicle VC.

The headquarters 120 analyzes data transmitted from the multiple vehicles VC. The headquarters 120 includes a CPU 122 , a ROM 124 , a storage device 126 , a peripheral circuit 127 and a communication device 129 that communicate with each other over a local area network 128 to be able to communicate.

The 10 FIG. 13 shows a procedure of a process related to the detection of misfire according to the present embodiment. The one in (a) in the 10 The process shown is done by the CPU 72 implemented that in the 9 misfire subroutine shown 74b executes that in ROM 74 is stored. Furthermore, the in (b) in the 10 process shown by the CPU 122 implemented the main misfire program 124a executes that in ROM 124 is stored. In the 10 processes that correspond to the 5 processes shown are labeled with the same step number for convenience. The Indian 10 The process shown will be described hereinafter along the timing of the misfire detection process.

This means that the CPU in the vehicle VC records 72 when in the process of S14 that appears in (a) in the 10 is shown, a confirmatory determination is made, the short rotation times T30 (0), T30 (6), T30 (12), T30 (18), T30 (24), T30 (30), T30 (36), T30 (42), T30 (48) ( S40 ). These short rotation times T30 constitute a rotation waveform variable that is a variable including information related to differences between short rotation times T30 at different angular intervals. Specifically, the short rotation time T30 is a period of time required for an angular interval from 30 ° CA before the compression top dead center to the compression top dead center, and it is a value which is nine times the timing of the occurrence of the compression top dead center. Therefore, the set data of the short rotation times T30 is a variable that indicates information on differences between the short rotation times T30 corresponding to compression top dead centers that are different from each other. The nine short rotation times T30 are all micro rotation times T30 used when calculating the inter-cylinder variable ΔTb (2) and the fluctuation pattern variables FL [02], FL [12], FL [32].

Next, the CPU performs 72 a process ( S42 ) to find the rotation speed NE and the charge efficiency η, and further performs a process ( S42 ) to determine the efficiency reduction scope vef, the target value Af * and the overlap scope RO in the period of the warm-up process ( S42 ). Then the CPU operates 72 the communication facility 79 to get those in the processes of S40 and S42 recorded data and the information (information on the existence of the execution) whether it is the data on the period of execution of the warm-up process or not, together with the identification information (vehicle ID) of the vehicle VC to the control center 120 transferred to ( S44 ).

The CPU 122 the headquarters 120 receives the transmitted data, as in (b) in the 10 will be shown ( S50 ). Then the CPU resets 122 the value of the variable obtained by the process of S50 has been recorded, instead of the input variables x (1) to x (11) ( S52 ). That is, the CPU 122 sets the short rotation time T30 (0) in place of the input variable x (1), sets the short rotation time T30 (6) in the place of the input variable x (2), puts the short rotation time T30 (12) in the place of the input variable x (3) and sets the short rotation time T30 (18) in place of the input variable x (4). The CPU 122 also sets the short rotation time T30 (24) in place of the input variable x (5), sets the short rotation time T30 (30) in the place of the input variable x (6) and sets the short rotation time T30 (36) in the place of the input variable x (7) ). Furthermore, the CPU 122 the short rotation time T30 (42) in the place of the input variable x (8) and sets the short rotation time T30 (48) in the place of the input variable x (9). In addition, the CPU 122 the rotational speed NE in the place of the input variable x (10) and the charge efficiency η in the place of the input variable x (11).

Next, the CPU determines 72 On the basis of the information on the existence of the execution, whether the recorded data is determined if the warm-up process is not executed or not ( S54 ). When it is determined that the acquired data has been obtained when the process was not executed (S54: YES), the CPU gives 122 the mapping created by the post-warmup mapping data 126a which is defined in the in the 9 storage device shown 126 are stored in the input variables x (1) to x (11) to get the value of the Calculate misfire variable PR, which is the output value of the mapping ( P.56 ).

In the present embodiment, the mapping is formed by a neural network in which the number of intermediate layers is “α”, the activation functions h1 to hα of each intermediate layer are ReLU and the activation function of the output layer is a softmax function. For example, the value of each node in the first intermediate layer is calculated by inputting the output when inputting the input variables x (1) to x (11) into the linear mapping represented by the coefficients wA (1) ji (j = 0 to n1 , i = 0 to 11) is generated in the activation function h1. That is, if m = 1, 2, ..., α, the value of each node of the m-th intermediate layer is determined by inputting the output of the linear mapping defined by the coefficient wA (m) into the activation function hm generated. In the 10 n1, n2, ..., nα are the numbers of nodes in the first, second, ... α. Intermediate layer. Here, wA (1) j0 and the like are bias parameters, and the input variable x (0) is defined as “1”.

If it is determined that the recorded data was determined during the period of the warm-up process ( S54 : NO), resets the CPU 122 the efficiency reduction scope vef in place of the input variable x (12), sets the target value Af * in the place of the input variable x (13) and sets the overlap extent RO in the place of the input variable x (14) ( S58 ).

Next, the CPU calculates 122 the value of the misfire variable PR, which is the output value of the mapping, by entering the input variables x (1) through x (14) into the mapping generated by the warm-up mapping data 126b which is defined in the in the 9 storage device shown 126 are stored ( S60 ).

In the present embodiment, the mapping is formed by a neural network in which the number of intermediate layers is “α”, the activation functions h1 to hα of each intermediate layer are ReLU and the activation function of the output layer is a softmax function. For example, the value of each node in the first intermediate layer is obtained by entering the outputs when the input variables x (1) to x (14) are entered into the linear mapping represented by the coefficient wB (1) ji (j = 0 to n1 , i = 0 to 14) is generated in the activation function h1. That is, if m = 1, 2, ..., α, the value of each node of the m-th intermediate layer is generated by inputting the output of the linear mapping defined by the coefficient wB (m) to the activation function . Here n1, n2, ..., nα are the number of nodes in the first, second, ... α. Intermediate layer. Here, wB (1) j0 and the like are bias parameters, and the input variable x (0) is defined as “1”.

Next, the CPU operates 122 the communication facility 129 to transmit a signal indicative of the value of the misfire variable PR to the vehicle VC to which the vehicle VC received by the process of S50 received data are transmitted ( S62 ), and first of all ends the sequence of processes described in (b) in the 10 to be shown. As in (a) in the 10 is shown, the CPU receives 72 the value of the misfire variable PR ( S46 ) and updates the processes S28 out.

Thus, in the present embodiment, the processes of P.56 and S60 in the main office 120 executed so that the processing load on the CPU 72 can be reduced.

Correspondence relationship

1. [1] Die Fehlzündungsdetektionseinrichtung entspricht der Steuerung 70. Die Ausführungseinrichtung, das heißt, die Verarbeitungsschaltkreise, entspricht der CPU 72 und dem ROM 74. Die Speichereinrichtung entspricht der Speichereinrichtung 76. Die Rotationskurvenformvariable entspricht der Zwischenzylindervariable ΔTb(2) und den Schwankungsmustervariablen FL[02], FL[12], FL[32]. Der Erfassungsprozess entspricht den Prozessen von S18 bis S22 und den Prozessen von S18, S20, S22a, der Bestimmungsprozess entspricht den Prozessen von S24 bis S36 und den Prozessen von S24a, S26a, S28 bis S36, und der Handhabungsprozess entspricht dem Prozess von S38. In der 2 entspricht der Warmlaufprozess dem Warmlaufkorrekturprozess M24, dem Zünd-Timing-Korrekturprozess M20, dem Zündvorgangsprozess M22, dem Ziel-Einlassphasendifferenzberechnungsprozess M32, dem Einlassphasendifferenzsteuerungsprozess M34, dem Zielwerteinstellprozess M46 und dem Einspritzventilbetätigungsprozess M44. In der 7 entspricht er dem Amplitudenwertvariablenausgabeprozess M50, dem Korrekturkoeffizientenberechnungsprozess M52, dem Dither-Korrekturprozess M54, dem Multiplikationsprozess M56, dem Korrekturkoeffizientenberechnungsprozess M58, dem Dither-Korrekturprozess M60 und dem Einspritzventilbetätigungsprozess M44, wenn die Amplitudenwertvariable α nicht null ist. Der Auswahlprozess entspricht dem Prozess von S8. Die Momentangeschwindigkeitsvariable entspricht der Kurzrotationszeit T30.
2. [2] Die Warmlaufbetriebsumfangsvariable entspricht dem Effizienzreduzierungsumfang vef, dem Zielwert Af*, dem Überlappungsumfang RO und der Amplitudenwertvariablen α.
3. [3] Die zum Umfang der Spätverstellung in Beziehung stehende Variable entspricht dem Effizienzreduzierungsumfang vef.
4. [4] Die Ventilspezifikationsvariableneinrichtung entspricht der Ventil-Timing-Einrichtung 40. Die Ventilkennlinienvariable entspricht dem Überlappungsumfang RO.
5. [5] Die Luft-Kraftstoff-Verhältnisvariable entspricht dem Zielwert Af*.
6. [6] Die Betriebspunktvariable entspricht der Rotationsgeschwindigkeit NE und der Ladungseffizienz η.
7. [7] Die erste Ausführungseinrichtung entspricht der CPU 72 und dem ROM 74. Die zweite Ausführungseinrichtung entspricht der CPU 122 und dem ROM 124. Die Rotationskurvenformvariable entspricht den Kurzrotationszeiten T30(0), T30(6), T30(12), ..., T30(48). Der Erfassungsprozess entspricht den Prozessen von S40 und S42, der fahrzeugseitige Übertragungsprozess entspricht dem Prozess von S44, und der fahrzeugseitige Empfangsprozess entspricht dem Prozess von S46. Der externe Empfangsprozess entspricht dem Prozess von S50, der Ausgabewertberechnungsprozess entspricht den Prozessen von S52 bis S60, und der externe Übertragungsprozess entspricht dem Prozess von S62.
8. [8] Das Datenanalysegerät entspricht der Zentrale 120.
9. [9] Die Steuerung für den Verbrennungsmotor entspricht der in der 9 gezeigten Steuerung 70.
10. [10] Der Computer entspricht der CPU 72 und dem ROM 74, und der CPU 72, der CPU 122, dem ROM 74 und dem ROM 124.

The correspondence relationship between the matters in the embodiment described above and the matters described in the “Outline of the Invention” section is as follows. Here, the correspondence relationship for each aspect number of the aspect described in the “Summary of the Invention” section is shown below.

1. [1] The misfire detection means corresponds to the controller 70 . The execution means, that is, the processing circuitry, corresponds to the CPU 72 and the ROM 74 . The storage device corresponds to the storage device 76 . The rotation waveform variable corresponds to the inter-cylinder variable ΔTb (2) and the fluctuation pattern variables FL [02], FL [12], FL [32]. The acquisition process corresponds to the processes of S18 to S22 and the processes of S18 , S20 , S22a , the determination process corresponds to the processes of S24 to S36 and the processes of S24a , S26a , S28 to S36 , and the handling process corresponds to the process of S38 . In the 2 the warm-up process corresponds to the warm-up correction process M24 , the ignition timing correction process M20 , the ignition process M22 , the target inlet phase difference calculation process M32 , the intake phase difference control process M34 , the target value setting process M46 and the injector actuation process M44 . In the 7th it corresponds to the amplitude value variable output process M50 , the correction coefficient calculation process M52 , the dither correction process M54 , the multiplication process M56 , the correction coefficient calculation process M58 , the dither correction process M60 and the injector actuation process M44 if the amplitude value variable α is not zero. The selection process corresponds to the process of S8 . The instantaneous speed variable corresponds to the short rotation time T30.
2. [2] The warm-up operation amount variable corresponds to the efficiency reduction amount vef, the target value Af *, the overlap amount RO and the amplitude value variable α.
3. [3] The variable related to the amount of retardation corresponds to the amount of efficiency reduction vef.
4. [4] The valve specification variable means corresponds to the valve timing means 40 . The valve characteristic variable corresponds to the extent of overlap RO.
5. [5] The air-fuel ratio variable corresponds to the target value Af *.
6. [6] The operating point variable corresponds to the rotation speed NE and the charge efficiency η.
7. [7] The first execution means corresponds to the CPU 72 and the ROM 74 . The second execution means corresponds to the CPU 122 and the ROM 124 . The rotation curve shape variable corresponds to the short rotation times T30 (0), T30 (6), T30 (12), ..., T30 (48). The acquisition process corresponds to the processes of S40 and S42 , the in-vehicle transfer process corresponds to the process of S44 , and the in-vehicle reception process corresponds to the process of S46. The external reception process corresponds to the process of S50 , the output value calculation process is the same as the processes of S52 to S60 , and the external transfer process is the same as the process of S62 .
8. [8] The data analyzer corresponds to the control center 120 .
9. [9] The control for the internal combustion engine corresponds to that in 9 control shown 70 .
10. [10] The computer corresponds to the CPU 72 and the ROM 74 , and the CPU 72 , the CPU 122 , the ROM 74 and the ROM 124 .

Other embodiments

The present embodiment described above can be modified and implemented as described below. The present embodiment and the following modified examples can be combined with each other within a range in which they do not contradict each other.

Valve characteristic variable

In the embodiment described above, the overlap amount RO is exemplified as the valve characteristic variable, but the present disclosure is not limited to such a manner. For example, the target inlet phase difference DIN * or the inlet phase difference DIN can be used. Furthermore, for example, the mean value of the target inlet phase difference DIN * and the inlet phase difference DIN in the execution cycle of the process from S26 and the like can be used.

Air-fuel ratio variable

In the embodiment described above, the target value Af * is shown as the air-fuel ratio variable by way of example, but the present disclosure is not limited in such a manner. For example, the mean value of the detection value Af in a predetermined period of time can be used.

Variable related to the amount of retardation of the ignition timing

In the embodiment described above, the amount of efficiency reduction is used as a variable related to the amount of retardation of the ignition timing, but the present disclosure is not limited in such a manner. For example, the mean value of the ignition timing correction amount Δaig in the execution timing of the process of FIG S26 and the like can be used.

For example, if the efficiency reduction amount vef is constant during the execution period of the warm-up process, a variable related to the amount of retardation of the ignition timing need not be included in the input to the mapping determined by the warm-up mapping data 76b and 126b is defined. However, it is not critical that the efficiency reduction amount vef is not included in the input to the mapping if the efficiency reduction amount vef is constant during the execution period of the warm-up process. By including the efficiency reduction scope vef in the input in the mapping in such a case, for example, several specifications with different efficiency reduction scope vef can be created with a single warm-up mapping data set 76b and 126b are processed.

Intermediate cylinder variable

The inter-cylinder variable ΔTb is not based on the difference in the short rotation time T30 limited according to the top compression dead center between two cylinders, the top compression dead center being reached one after the other by 720 ° CA apart. For example, the inter-cylinder variable ΔTb may be a difference in the short rotation time T30 corresponding to top compression dead centers between cylinders that are 360 ° CA apart in terms of reaching time of 720 ° CA top compression dead centers. In this case the intermediate cylinder variable ΔTb (2) is “T30 (12) -T30 (24) - {T30 (36) -T30 (48)}”.

Instead of the difference between the values of the difference between the short rotation times T30, which are separated by 720 ° CA, corresponding to the top compression dead centers of two cylinders, the difference in the short rotation time T30 can also be the top compression dead centers of the cylinder at which the misfire is to be detected and of another cylinder correspond.

For example, the intermediate cylinder variable may further be a ratio between the short rotation times T30 corresponding to the top compression dead centers of the two cylinders.

The short rotation time in defining the inter-cylinder variable ΔTb is not limited to the time required for the rotation of 30 ° CA, but may be, for example, a time required for the rotation of 45 ° CA. In this period of time, the short rotation time can be a period of time that is required for the rotation of an angular interval that is less than or equal to the interval at which the compression top dead center is reached. The compression top dead center reaching interval means the interval between the rotation angles of the crankshaft 24 at which the top compression dead center is reached. An angular interval that is smaller than or equal to the interval when the top compression dead center is reached can also be referred to as a very short angular interval. The speed of rotation of the crankshaft 24 in each of the several very short angular intervals can be referred to as an instantaneous speed.

In the description given above, instead of the short rotation time, an instantaneous rotation speed which is determined by dividing the predetermined angular interval by the time required for the rotation of the predetermined angular interval can also be used.

Fluctuation pattern variable

The definition of the fluctuation pattern variable is not limited to that exemplified in the embodiment described above. For example, the definition of the fluctuation pattern variable can be changed by changing the inter-cylinder variable ΔTb to that shown in the section “Inter-cylinder variable” as an example.

Furthermore, it is not essential to define the fluctuation pattern variable as a ratio between the inter-cylinder variables ΔTb corresponding to the timings of occurrence of the different compression top dead centers, and a difference may be used in place of the ratio. Even in this case, the value of the misfire variable PR can be calculated by reflecting the fact that the fluctuation pattern variable is changed according to the operating point by changing the operating point variable of the internal combustion engine 10 is included in the input.

Rotation waveform variable

In the process of S26 the rotational waveform variable is formed by the inter-cylinder variable ΔTb (2) and the fluctuation pattern variables FL [02], FL [12], FL [32]. However, this is not a limitation. For example, the fluctuation pattern variable that constitutes the rotation waveform variable may be one or two of the fluctuation pattern variables FL [02], FL [12], FL [32]. Furthermore, four or more fluctuation pattern variables such as fluctuation pattern variables FL [02], FL [12], FL [32], FL [42] and the like may be included.

In the processes of P.56 and S60 For example, the rotation waveform variable is formed by the short rotation time T30 corresponding to each of the nine timings at which the timing of the occurrence of the compression top dead center are different from each other, however, the present disclosure is not limited to such a manner. With the compression top dead center of a cylinder at which misfire is to be detected as the center, for example, the rotation waveform variable is formed by the short rotation time T30 in each of the sections obtained by dividing a zone of the two or more angular interval in which the compression top dead center occurs, can be determined through an interval of 30 ° CA. In the above description, it is not critical to set the compression top dead center of the cylinder at which misfire is to be detected as the center. Furthermore, the short rotation time is not limited to the period of time that is required for the rotation of an interval of 30 ° CA. Instead of the short rotation time, an instantaneous rotation speed can also be determined by dividing the predetermined angular interval through the time period is divided, which is necessary for the rotation of the predetermined angular interval.

Operating point variable

The operating point variable is not limited to the rotation speed NE and the charging efficiency η. For example, an intake air amount Ga and the rotation speed NE can be used. Further, for example, the injection amount and the rotation speed NE can be used when a compression ignition type internal combustion engine is used, as described in the section “Internal combustion engine” below. It is not critical to use the operating point variable as an input to the mapping. When applied to an internal combustion engine mounted in a series hybrid vehicle described in the "Vehicle" section below, when the internal combustion engine is operated only at a specific operating point and the like, the value of the misfire variable PR can be calculated with high accuracy without including the operating point variable in the input variable.

External transfer process

In the process of S62 the value of the misfire variable PR is transmitted, but the present disclosure is not limited in such a manner. For example, the values of the prototype variables yR (1) and yR (2) can be transferred. Furthermore, the headquarters 120 for example the process of S28 to S36 and the result of determination of whether there is an abnormality can be transmitted.

Handling process

In the above-described embodiment, the occurrence of misfires is detected via visual information by operating the warning lamp 100 reported. However, this is not a limitation. For example, the occurrence of misfires can be reported by acoustic information by operating a loudspeaker. Furthermore, the 1 illustrated control 70 the communication facility 79 include, and the communication device 79 can be operated to transmit a signal indicating that a misfire has occurred to the user's portable terminal. This can be implemented in that an application program for carrying out the notification process is installed in the user's portable terminal.

The handling process is not limited to the notification process. For example, an operating process can be used that operates an operating unit to cause the combustion of the air-fuel mixture in the combustion chamber 18th of the internal combustion engine 10 in accordance with information indicating that a misfire has occurred. Specifically, for example, the ignition timing of a cylinder in which a misfire has occurred can be made with the operating unit as the igniter 22nd can be adjusted early. Further, for example, the fuel injection amount for the cylinder in which the misfire has occurred can be determined with the operation unit as the fuel injection valve 20th increase.

Input in the mapping

The input to the neural network, the input to the regression equation described in the “Machine learning algorithm” section, and the like are not limited to those in which each quantity is formed by a single physical quantity or the fluctuation pattern variable FL. With respect to some of the plural types of physical quantities and the fluctuation pattern variables FL used as the input to the mapping in the above-described embodiment and the like, for example, some principal components obtained through their principal component analysis can be used as the direct input into the neural network and the regression equation can be used. If the main component is an input to the neural network and the regression equation, it is not critical that only part of the input to the neural network and the regression equation is the main component, and all of the input can be the main component. If the major component is an input to the mapping, the post-warmup includes mapping data 76a and 126a and the warm-up mapping data 76b and 126b Data defining a mapping that determines the main component.

Mapping data or mapping data or map data

The mapping data that defines the mapping that is used for the calculation performed in the vehicle can be the post-warm-up mapping data 126a and the warm-up mapping data 126b be.

The mapping data that define the mapping that is used for the headquarters 120 The calculation performed is used, the post-warm-up mapping data 76a and the warm-up mapping data 76b be.

For example, according to the description of the 10 the number of intermediate layers in the neural network expressed as more than two layers. However, this is not a limitation.

In the embodiment described above, the activation functions are h, h1, h2, ... hα ReLU, and the activation function of the output is a softmax function, but the present disclosure is not limited to such a manner. For example, the activation functions h, h1, h2, ... hα can be hyperbolic tangent functions. For example, the activation functions h, h1, h2, ... hα can be logistic sigmoid functions.

For example, the activation function of the output can be a logistic sigmoid function. In this case, for example, the number of nodes in the output layer can be one and the output variable can be the misfire variable PR. In this case, the presence or absence of an abnormality can be determined by determining as abnormal when the value of the output variable is greater than or equal to a predetermined value.

Machine learning algorithm

The machine learning algorithm is not limited to using a neural network. For example, a regression equation can be used. This corresponds to the neural network, which does not include an intermediate layer. A support vector machine can also be used, for example. In this case, the value of the output itself does not matter, and whether or not the misfire has occurred is appropriately expressed by whether or not the value is positive. In other words, it is different from when the value of the misfire variable has a value of 3 or more, and these values represent the probability of misfire.

Learning step

In the embodiment described above, the learning is performed in a situation where misfire accidentally occurs, but the present disclosure is not limited to such a manner. For example, the learning can be carried out in a situation where misfire occurs continuously in a specific cylinder. However, in this case, the inter-cylinder variable ΔTb used for the inter-cylinder variable and the fluctuation pattern variable which will be the input to the mapping can be the difference between the short rotation times T30 corresponding to the top compression dead centers of the cylinder at which misfire is to be detected and other cylinders as described in the section "Intermediate cylinder variable".

Data analyzer

The process of (b) in the 10 can for example be performed by a portable terminal owned by the user. This can be implemented by using an application program to execute the process of (b) in the 10 is installed in the portable device. At this time, for example, the transmission / reception process for the vehicle ID can be canceled by changing the effective distance of the data transmission in the process of S44 is set to approximately the length of the vehicle and the like.

Execution facility

The executing device is not limited to a device which is the CPU 72 ( 122 ) and the ROM 74 ( 124 ) and executes the software process. For example, dedicated hardware circuitry (e.g., ASIC, etc.) can be arranged to process at least part of the software process carried out in the embodiment described above. In other words, the execution device only needs to have one of the configurations (a) to (c) below. (a) A processing device that executes all of the above-mentioned processes according to a program and a program storage device such as a ROM that stores the program are provided. (b) A processing device and a program storage device which execute part of the above-mentioned processes according to a program, and a dedicated hardware circuit which executes the remaining processes are provided. (c) Dedicated hardware circuitry that performs all of the above processes is provided. The software execution device, which comprises the processing device and the program storage device, or the dedicated hardware circuit can be provided in the plurality here. That is, the above processes can be performed by processing circuitry including at least one of one or more software executing devices and one or more dedicated hardware circuits. The program storage device, that is, a computer-readable medium, comprises various usable media that can be accessed with a general-purpose or a dedicated computer.

Storage facility

In the embodiment described above, the storage means are for storing the post-warm-up mapping data 76a , 126a and the warm-up mapping data 76b , 126b and the storage device (ROM 74 , 124 ) to save the misfire program 74a and the misfire main program 124a separate storage facilities. However, this is not a limitation.

computer

The computer is not limited to a computer that has an executing device such as the CPU 72 and the ROM 74 that are mounted in the vehicle and an execution device such as the CPU 122 and the ROM 124 that are in the headquarters 120 provided includes. For example, the computer can be from an execution device that is mounted in the vehicle and the execution device that is in the control center 120 is provided, and an execution device such as the CPU and the ROM in a portable terminal of a user. This can be implemented, for example, by the process of S62 in the 10 is performed as a process that transmits to the user's portable terminal, and the processes of S46, S28 to S38 in (a) in the 10 be executed on the portable device. In particular, an in-vehicle execution device that is executed by the CPU 72 and the ROM 74 is designed, be designed not to carry out the vehicle-side receiving process and the handling process. A reception execution device, which is included in the portable terminal, can be designed to execute at least the vehicle-side reception process.

Internal combustion engine

In the embodiment described above, the injector is in the cylinder, the fuel in the combustion chamber 18th injected, shown by way of example as the fuel injection valve, but is not limited thereto. For example, the fuel injector may be an intake port injector that delivers fuel to the intake port 12th injects. Furthermore, for example, both an intake port injection valve and an injection valve can be provided in the cylinder.

The internal combustion engine is not limited to a spark ignition type internal combustion engine and may be a compression ignition type internal combustion engine using light oil or the like as fuel.

vehicle

The vehicle is not limited to a series / parallel hybrid vehicle. For example, the vehicle may be a parallel hybrid vehicle or a series hybrid vehicle. Since it is not limited to the hybrid vehicle, it may be a vehicle in which a device that generates the thrust of the vehicle is only an internal combustion engine.

Claims (11)

A misfire detection device for an internal combustion engine, the misfire detection device comprising: memory means; and processing circuits, the storage device storing first mapping data corresponding to the case that a warm-up process is carried out in a catalytic converter which is arranged in an exhaust passage of an internal combustion engine, and second mapping data corresponding to the case that the warm-up process is not carried out, wherein each of the first and second mapping data defines a mapping in which a misfire variable is output using a rotation waveform variable, the misfire variable being a variable related to a probability of occurrence of misfire, the processing circuitry is configured as follows execute: a detection process that detects the rotation waveform variable based on a detection value of a sensor configured to detect a rotational behavior of a crankshaft of the internal combustion engine; a determination process that determines whether the misfire is occurring based on an output of the mapping using the variable as input acquired by the acquisition process, a handling process that handles the occurrence of misfire in the case that is determined in the determination process by operating predetermined hardware that the misfire has occurred, and a selection process that selects either the first mapping data or the second mapping data used in the determination process in accordance with whether the warm-up process has been performed, with an interval between the angles at which upper compression dead centers can be reached in the internal combustion engine, a reach interval is several angle intervals that are smaller than the reach interval, several very small angle intervals are, a rotational speed of the crankshaft at each of the several very small angular intervals is an instantaneous speed and a variable which is related to the instantaneous speed is an instantaneous speed variable, the rotation waveform variable is a variable indicating a difference between a plurality of values of an instantaneous speed variable or corresponding to a plurality of different very short angular intervals, and the mapping returns a value of the misfire variable by performing a join operation on the value of the rotation waveform variable and a parameter learned through machine learning. Misfire detection device according to Claim 1 wherein the input of the mapping defined by the first mapping data when the warm-up process is executed includes a warm-up operation amount variable that is a variable that is related to an operation amount of an operating unit of the internal combustion engine via the warm-up process, the detection process includes a process that detects the warm-up operation amount variable when the warm-up process is performed, and the determination process includes a process that determines the presence of the misfire based on the output of the mapping that further uses the warm-up operation amount variable that is detected by the detection process as the input, when the warm-up process is in progress. Misfire detection device according to Claim 2 , wherein the warm-up process includes a process that retards an ignition timing compared to when the warm-up process is not performed, and the warm-up operation amount variable detected by the detection process includes a variable that corresponds to an amount of retardation of the Firing timings is related. Misfire detection device according to Claim 2 or 3 wherein the internal combustion engine comprises valve characteristics variable means that allows valve characteristics of an intake valve to be changed, the warm-up process includes a process that operates the valve characteristic variable means, and the warm-up operation amount variable that is detected by the detection process includes a valve characteristic variable that is a variable that is is related to the valve characteristics. Misfire detection device according to one of the Claims 2 to 4th , wherein the warm-up process includes a process that changes an air-fuel ratio of an air-fuel mixture burned in a combustion chamber of the internal combustion engine according to a progress status of the warm-up process, and the warm-up operation amount variable detected by the detection process, a Includes air-fuel ratio variable, which is a variable related to air-fuel ratio. Misfire detection device according to one of the Claims 1 to 5 wherein the input of the mapping includes an operating point variable that is a variable that defines an operating point of the internal combustion engine, the detection process includes a process that detects the operating point variable, the determination process is a process that the presence of the misfire based on the output of the mapping which further uses the operating point variable detected by the detection process as the input, and the mapping outputs a value of the misfire variable by performing a join operation on the rotational waveform variable, the operating point variable and the machine learning learned parameter. A misfire detection system for an internal combustion engine, the misfire detection system comprising: the processing circuitry and the memory device of any one of Claims 1 to 5 wherein the determination process includes an output value calculation process of calculating an output value of the mapping using a variable detected by the detection process as the input, the processing circuitry including a first execution device and a second execution device, the first execution device at least partially mounted in the vehicle and is configured to carry out the acquisition process, an on-vehicle transmission process, the data acquired by the acquisition process to be transmitted outside the vehicle, an in-vehicle reception process that receives a signal based on a calculation result of the output value calculation process, and the handling process , and the second execution device is arranged outside of the vehicle and is designed to carry out an external reception process that receives data that is transmitted by the vehicle-side transmission process to execute the output value calculation process, the selection process, and an external transmission process that transmits a signal based on a calculation result of the output value calculation process to the vehicle. Data analyzer, which the second execution device and the storage device according to Claim 7 includes. Control for an internal combustion engine, which the first execution device according to Claim 7 includes. A misfire detection method for an internal combustion engine, wherein the detection process, the determination process and the handling process are according to any one of Claims 1 to 6 be carried out by a computer. Reception execution device in a misfire detection system for an internal combustion engine according to Claim 7 , wherein the first execution device comprises an in-vehicle execution device, which is mounted in a vehicle, and a reception execution device, which is separate from the in-vehicle execution device, the in-vehicle execution device is configured to carry out the acquisition process and an in-vehicle transmission process of the data generated by the Acquisition process are detected, transmitted to the outside of the vehicle, carried out and the reception execution device is provided in a portable terminal and is designed to carry out at least the vehicle-side reception process.

Images